Machine learning for antidepressant treatment selection in depression.

Finding the right antidepressant for the individual patient with major depressive disorder can be a difficult endeavor and is mostly based on trial-and-error. Machine learning (ML) is a promising tool to personalize antidepressant prescription. In this review, we summarize the current evidence of ML in the selection of antidepressants and conclude that its value for clinical practice is still limited. Apart from the current focus on effectiveness, several other factors should be taken into account to make ML-based prediction models useful for clinical application.

Transmural collaborative care model for the review of antipsychotics: a feasibility study of a complex intervention.

General practitioners (GPs) are often unaware of antipsychotic (AP)-induced cardiovascular risk (CVR) and therefore patients using atypical APs are not systematically monitored. We evaluated the feasibility of a complex intervention designed to review the use of APs and advise on CVR-lowering strategies in a transmural collaboration. A mixed methods prospective cohort study in three general practices in the Netherlands was conducted in 2021. The intervention comprised three steps: a digital information meeting, a multidisciplinary meeting, and a shared decision-making visit to the GP. We assessed patient recruitment and retention rates, advice given and adopted, and CVR with QRISK3 score and mental state with MHI-5 at baseline and three months post-intervention. GPs invited 57 of 146 eligible patients (39%), of whom 28 (19%) participated. The intervention was completed by 23 (82%) and follow-up by 18 participants (64%). At the multidisciplinary meeting, 22 (78%) patients were advised to change AP use. Other advice concerned medication (other than APs), lifestyle, monitoring, and psychotherapy. At 3-months post-intervention, 41% (28/68) of this advice was adopted. Our findings suggest that this complex intervention is feasible for evaluating health improvement in patients using AP in a trial.

A genetic risk score to predict treatment nonresponse in psychotic depression.

Psychotic depression is a severe and difficult-to-treat subtype of major depressive disorder for which higher rates of treatment-resistant depression were found. Studies have been performed aiming to predict treatment-resistant depression or treatment nonresponse. However, most of these studies excluded patients with psychotic depression. We created a genetic risk score (GRS) based on a large treatment-resistant depression genome-wide association study. We tested whether this GRS was associated with nonresponse, nonremission and the number of prior adequate antidepressant trials in patients with a psychotic depression. Using data from a randomized clinical trial with patients with a psychotic depression (n = 122), we created GRS deciles and calculated positive prediction values (PPV), negative predictive values (NPV) and odds ratios (OR). Nonresponse and nonremission were assessed after 7 weeks of treatment with venlafaxine, imipramine or venlafaxine plus quetiapine. The GRS was negatively correlated with treatment response (r = -0.32, p = 0.0023, n = 88) and remission (r = -0.31, p = 0.0037, n = 88), but was not correlated with the number of prior adequate antidepressant trials. For patients with a GRS in the top 10%, we observed a PPV of 100%, a NPV of 73.7% and an OR of 52.4 (p = 0.00072, n = 88) for nonresponse. For nonremission, a PPV of 100%, a NPV of 51.9% and an OR of 21.3 (p = 0.036, n = 88) was observed for patients with a GRS in the top 10%. Overall, an increased risk for nonresponse and nonremission was seen in patients with GRSs in the top 40%. Our results suggest that a treatment-resistant depression GRS is predictive of treatment nonresponse and nonremission in psychotic depression.